oalib
Search Results: 1 - 10 of 100 matches for " "
All listed articles are free for downloading (OA Articles)
Page 1 /100
Display every page Item
Accelerating Asymptotically Exact MCMC for Computationally Intensive Models via Local Approximations  [PDF]
Patrick R. Conrad,Youssef M. Marzouk,Natesh S. Pillai,Aaron Smith
Statistics , 2014,
Abstract: We construct a new framework for accelerating Markov chain Monte Carlo in posterior sampling problems where standard methods are limited by the computational cost of the likelihood, or of numerical models embedded therein. Our approach introduces local approximations of these models into the Metropolis-Hastings kernel, borrowing ideas from deterministic approximation theory, optimization, and experimental design. Previous efforts at integrating approximate models into inference typically sacrifice either the sampler's exactness or efficiency; our work seeks to address these limitations by exploiting useful convergence characteristics of local approximations. We prove the ergodicity of our approximate Markov chain, showing that it samples asymptotically from the \emph{exact} posterior distribution of interest. We describe variations of the algorithm that employ either local polynomial approximations or local Gaussian process regressors. Our theoretical results reinforce the key observation underlying this paper: when the likelihood has some \emph{local} regularity, the number of model evaluations per MCMC step can be greatly reduced without biasing the Monte Carlo average. Numerical experiments demonstrate multiple order-of-magnitude reductions in the number of forward model evaluations used in representative ODE and PDE inference problems, with both synthetic and real data.
Embarrassingly Parallel Variational Inference in Nonconjugate Models  [PDF]
Willie Neiswanger,Chong Wang,Eric Xing
Computer Science , 2015,
Abstract: We develop a parallel variational inference (VI) procedure for use in data-distributed settings, where each machine only has access to a subset of data and runs VI independently, without communicating with other machines. This type of "embarrassingly parallel" procedure has recently been developed for MCMC inference algorithms; however, in many cases it is not possible to directly extend this procedure to VI methods without requiring certain restrictive exponential family conditions on the form of the model. Furthermore, most existing (nonparallel) VI methods are restricted to use on conditionally conjugate models, which limits their applicability. To combat these issues, we make use of the recently proposed nonparametric VI to facilitate an embarrassingly parallel VI procedure that can be applied to a wider scope of models, including to nonconjugate models. We derive our embarrassingly parallel VI algorithm, analyze our method theoretically, and demonstrate our method empirically on a few nonconjugate models.
Optimization Monte Carlo: Efficient and Embarrassingly Parallel Likelihood-Free Inference  [PDF]
Edward Meeds,Max Welling
Computer Science , 2015,
Abstract: We describe an embarrassingly parallel, anytime Monte Carlo method for likelihood-free models. The algorithm starts with the view that the stochasticity of the pseudo-samples generated by the simulator can be controlled externally by a vector of random numbers u, in such a way that the outcome, knowing u, is deterministic. For each instantiation of u we run an optimization procedure to minimize the distance between summary statistics of the simulator and the data. After reweighing these samples using the prior and the Jacobian (accounting for the change of volume in transforming from the space of summary statistics to the space of parameters) we show that this weighted ensemble represents a Monte Carlo estimate of the posterior distribution. The procedure can be run embarrassingly parallel (each node handling one sample) and anytime (by allocating resources to the worst performing sample). The procedure is validated on six experiments.
A Tool for Programming Embarrassingly Task Parallel Applications on CoW and NoW  [PDF]
Patrizio Dazzi
Computer Science , 2013,
Abstract: Embarrassingly parallel problems can be split in parts that are characterized by a really low (or sometime absent) exchange of information during their computation in parallel. As a consequence they can be effectively computed in parallel exploiting commodity hardware, hence without particularly sophisticated interconnection networks. Basically, this means Clusters, Networks of Workstations and Desktops as well as Computational Clouds. Despite the simplicity of this computational model, it can be exploited to compute a quite large range of problems. This paper describes JJPF, a tool for developing task parallel applications based on Java and Jini that showed to be an effective and efficient solution in environment like Clusters and Networks of Workstations and Desktops.
Software Alchemy: Turning Complex Statistical Computations into Embarrassingly-Parallel Ones  [PDF]
Norman Matloff
Statistics , 2014,
Abstract: The growth in the use of computationally intensive statistical procedures, especially with Big Data, has necessitated the usage of parallel computation on diverse platforms such as multicore, GPU, clusters and clouds. However, slowdown due to interprocess communication costs typically limits such methods to "embarrassingly parallel" (EP) algorithms, especially on non-shared memory platforms. This paper develops a broadly-applicable method for converting many non-EP algorithms into statistically equivalent EP ones. The method is shown to yield excellent levels of speedup for a variety of statistical computations. It also overcomes certain problems of memory limitations.
Pretty Fast Analysis: An embarrassingly parallel algorithm for biological simulation analysis  [PDF]
David N. Lebard
Physics , 2008,
Abstract: A parallel code has been written in FORTRAN90, C, and MPI for the analysis of biological simulation data. Using a master/slave algorithm, the software operates on AMBER generated trajectory data using either UNIX or MPI file IO, and it supports up to 15 simultaneous function calls. This software has been performance tested on the Ranger Supercomputer on trajectory data of an aqueous bacterial reaction center micelle. Although the parallel reading is poor, the analysis algorithm itself shows embarrassingly parallel speedup up to 1024 compute nodes. At this CPU count the overall scaling of the software compares well NAMD's best reported speedup, and outperforms AMBER's best known scaling by a factor of 3, while using only a small number of function calls and a short trajectory length.
Parallel MCMC with Generalized Elliptical Slice Sampling  [PDF]
Robert Nishihara,Iain Murray,Ryan P. Adams
Statistics , 2012,
Abstract: Probabilistic models are conceptually powerful tools for finding structure in data, but their practical effectiveness is often limited by our ability to perform inference in them. Exact inference is frequently intractable, so approximate inference is often performed using Markov chain Monte Carlo (MCMC). To achieve the best possible results from MCMC, we want to efficiently simulate many steps of a rapidly mixing Markov chain which leaves the target distribution invariant. Of particular interest in this regard is how to take advantage of multi-core computing to speed up MCMC-based inference, both to improve mixing and to distribute the computational load. In this paper, we present a parallelizable Markov chain Monte Carlo algorithm for efficiently sampling from continuous probability distributions that can take advantage of hundreds of cores. This method shares information between parallel Markov chains to build a scale-mixture of Gaussians approximation to the density function of the target distribution. We combine this approximation with a recent method known as elliptical slice sampling to create a Markov chain with no step-size parameters that can mix rapidly without requiring gradient or curvature computations.
Parallelizing MCMC with Random Partition Trees  [PDF]
Xiangyu Wang,Fangjian Guo,Katherine A. Heller,David B. Dunson
Statistics , 2015,
Abstract: The modern scale of data has brought new challenges to Bayesian inference. In particular, conventional MCMC algorithms are computationally very expensive for large data sets. A promising approach to solve this problem is embarrassingly parallel MCMC (EP-MCMC), which first partitions the data into multiple subsets and runs independent sampling algorithms on each subset. The subset posterior draws are then aggregated via some combining rules to obtain the final approximation. Existing EP-MCMC algorithms are limited by approximation accuracy and difficulty in resampling. In this article, we propose a new EP-MCMC algorithm PART that solves these problems. The new algorithm applies random partition trees to combine the subset posterior draws, which is distribution-free, easy to resample from and can adapt to multiple scales. We provide theoretical justification and extensive experiments illustrating empirical performance.
CRBLASTER: A Parallel-Processing Computational Framework for Embarrassingly-Parallel Image-Analysis Algorithms  [PDF]
Kenneth John Mighell
Physics , 2010, DOI: 10.1086/656566
Abstract: The development of parallel-processing image-analysis codes is generally a challenging task that requires complicated choreography of interprocessor communications. If, however, the image-analysis algorithm is embarrassingly parallel, then the development of a parallel-processing implementation of that algorithm can be a much easier task to accomplish because, by definition, there is little need for communication between the compute processes. I describe the design, implementation, and performance of a parallel-processing image-analysis application, called CRBLASTER, which does cosmic-ray rejection of CCD (charge-coupled device) images using the embarrassingly-parallel L.A.COSMIC algorithm. CRBLASTER is written in C using the high-performance computing industry standard Message Passing Interface (MPI) library. The code has been designed to be used by research scientists who are familiar with C as a parallel-processing computational framework that enables the easy development of parallel-processing image-analysis programs based on embarrassingly-parallel algorithms. The CRBLASTER source code is freely available at the official application website at the National Optical Astronomy Observatory. Removing cosmic rays from a single 800x800 pixel Hubble Space Telescope WFPC2 image takes 44 seconds with the IRAF script lacos_im.cl running on a single core of an Apple Mac Pro computer with two 2.8-GHz quad-core Intel Xeon processors. CRBLASTER is 7.4 times faster processing the same image on a single core on the same machine. Processing the same image with CRBLASTER simultaneously on all 8 cores of the same machine takes 0.875 seconds -- which is a speedup factor of 50.3 times faster than the IRAF script. A detailed analysis is presented of the performance of CRBLASTER using between 1 and 57 processors on a low-power Tilera 700-MHz 64-core TILE64 processor.
Orthogonal parallel MCMC methods for sampling and optimization  [PDF]
L. Martino,V. Elvira,D. Luengo,J. Corander,F. Louzada
Statistics , 2015,
Abstract: Monte Carlo (MC) methods are widely used in statistics, signal processing and machine learning. A well-known class of MC methods are Markov Chain Monte Carlo (MCMC) algorithms. In order to foster better exploration of the state space, specially in high-dimensional applications, several schemes employing multiple parallel MCMC chains have been recently introduced. In this work, we describe a novel parallel interacting MCMC scheme, called orthogonal MCMC (O-MCMC), where a set of vertical parallel MCMC chains share information using some horizontal MCMC techniques working on the entire population of current states. More specifically, the vertical chains are led by random-walk proposals, whereas the horizontal MCMC techniques employ independent proposals, thus allowing an efficient combination of global exploration and local approximation. The interaction is contained in these horizontal iterations. Within the analysis of different implementations of O-MCMC, novel schemes for reducing the overall computational cost of parallel multiple try Metropolis (MTM) chains are also presented. Furthermore, a modified version of O-MCMC for optimization is provided by considering parallel simulated annealing (SA) algorithms. Finally, we also discuss the application of O-MCMC in a big bata framework. Numerical results show the advantages of the proposed sampling scheme in terms of efficiency in the estimation, as well as robustness in terms of independence with respect to initial values and the choice of the parameters.
Page 1 /100
Display every page Item


Home
Copyright © 2008-2017 Open Access Library. All rights reserved.